In this prospective study, data of 26 patients with bilateral hip OA and 26 gender, age and walking speed matched patients with unilateral hip OA were included. All patients were scheduled for a unilateral THA on the more affected/more painful side. Furthermore, 26 healthy controls, also matched for gender, age and walking speed, were selected from our reference database and included for comparison (Table 1). Healthy controls were included if they had no history of orthopedic surgeries, chronic or neuromuscular diseases or recent injuries to the lower extremities. None of the enrolled healthy controls reported ankle, knee, hip or back pain at the time of measurement. All patients and healthy controls gave written informed consent prior to participation in the original studies. Our institution’s medical ethics committee approved the studies under the numbers 319/11, 122/14 and 497/15.

Patients underwent a three-dimensional gait analysis in the week before surgery, whereas the healthy controls visited the gait laboratory within the framework of a reference measurement. The healthy controls were asked to walk at a walking speed of approximately 1.0 ms⁻¹, comparable to patients shortly before surgery (Schmidt et al., 2017). Kinematic data were collected using 8 Vicon MX T10 cameras (VICON Motion Systems, Oxford, United Kingdom) operating at 200 Hz and two AMTI force plates (Advanced Mechanical Technology Inc., Watertown, MA, United States) were used to synchronously collect ground reaction forces at 1000 Hz. A marker set with markers on the medial malleolus, medial femoral condyle and greater trochanter (Stief et al., 2013) additional to the standardized Plug-in-Gait marker set (Kadaba et al., 1990) was used.

Vicon Nexus (version 2.10) was used to process the marker trajectories. The resulting marker trajectories were filtered using a Woltring filter with a mean squared error setting of 10 (Woltring, 1986). The Plug-in-Gait model (VICON Motion Systems, Oxford, United Kingdom) is Vicon’s implementation of the Conventional Gait Model and is still widely used in the clinical gait analysis community (Baker et al., 2018). This model was used to calculate the kinematic and kinetic gait variables. The hip joint centers were obtained according to a geometrical prediction method using regression equations (Davis et al., 1991) whereas the centers of rotation for the ankle and knee joints were defined statically as the midpoint between the medial and lateral malleolus and femoral condyle markers. Kinematic outputs were calculated from the embedded coordinate system information. The joint angles mainly describe the orientation of the distal segment with respect to that of the proximal segment. The orientation of the pelvis is output as segment angles with respect to the laboratory-based axis system, as is the transverse plane alignment of the foot (the so-called foot progression angle). Inverse dynamics were used to estimate joint moments from force plate measurements of the ground reaction, an estimate of segment accelerations from kinematic data and estimates of segment inertial parameters (Kadaba et al., 1990; Davis et al., 1991).

For all patients the leg to be operated on (the more affected/more painful side) was considered the ipsilateral leg, whereas the non-affected side (unilateral patients) and the less affected side (bilateral patients) was called the contralateral leg. For the healthy controls the kinematics and kinetics of the left side were chosen randomly for comparison.

Kinematic and kinetic data were exported to Matlab (version R2020b, The Mathworks Inc., Natick, MA, United States) to normalize the data over the gait cycle and average the parameters over five good trials.

The following kinematic gait parameters, which are typically influenced by hip OA, were included for further analysis: thorax tilt and lean (the lateral displacement of the trunk relative to the supporting limb), pelvic tilt and obliquity, hip flexion-extension, knee flexion-extension, ankle dorsiflexion-plantarflexion and the foot progression angle (the angle of the long axis of the foot segment relative to the direction of walking). External ankle, knee and hip joint moments (expressed in Nmkg⁻¹) were defined as kinetic outcomes. In addition, the impulse for the knee adduction and hip adduction moment (area under the curve (Herger et al., 2019)) were calculated.

For the peak external knee and hip adduction moment during the first and second half of the stance phase as well as for the impulse of the adduction moments, an absolute symmetry index (ASI) was determined (Robinson et al., 1987; Bosch and Rosenbaum, 2010) by the following equation: A S I =   | 2 ( X L −   X R ) ( X L +   X R ) |   × 100 %

X_L and X_R are the parameters of the ipsilateral and contralateral leg for patients and the left and right leg for the healthy controls. An ASI of 0 indicates perfect symmetry.

Shapiro-Wilk tests were used to test for normal distribution of the anthropometrics. Since the data showed a normal distribution, one-way ANOVA (group) with Bonferroni posthoc tests were used to determine statistical differences between the anthropometrics and walking speed of the patient groups and healthy controls.

The selected gait kinematics and kinetics were evaluated using Statistical Parametric Mapping (SPM). SPM originates from the Random Field Theory (Adler and Taylor, 2007) and has been validated for 1D biomechanical data by Pataky et al. (Pataky et al., 2013). For the kinematics and kinetics, normal distribution was tested and confirmed by SPM. The SPM analyses were performed in MATLAB using the open-source spm1d code (version M.0.4.3) with a general critical threshold (p-value) of α = 0.05. SPM paired sample t-tests were used to compare the ipsilateral and contralateral legs of the patients. Data of the bilateral, unilateral and healthy controls were compared first with an SPM ANOVA and afterwards with independent two-sample t-tests (critical p-value corrected for multiple testing, α = 0.05/3). Data were considered significantly different when the waveform exceeded this critical threshold for more than four successive time points, i.e., at least 4% of the gait cycle (Wesseling et al., 2018).

Symmetry between patients and healthy controls was tested with Kruskal-Wallis tests and Mann-Whitney tests as the data appeared not be to be normally distributed. P-values ≤ 0.05 were considered significant for all analyses.

In all groups (bilateral patients, unilateral patients and healthy controls), 16 women and 10 men were included. As patients and healthy controls were age and speed matched, no significant differences were present in age and speed, nor in weight and height (Table 1). Only the Body Mass Index (BMI) was significantly higher in unilateral hip OA patients compared to healthy controls. All bilateral patients had a Kellgren-Lawrence score (Kellgren and Lawrence, 1957) of at least 5 in both legs (Grade 3), whereas the unilateral patients were selected not to have a score exceeding 2 (Grade 1) in the contralateral leg. All patients reported significant pain in the ipsilateral hip joint: 34 patients scored 45.9 (11.4) on the pain section of the Hip disability and Osteoarthritis Outcome Score (Nilsdotter et al., 2003; Blasimann et al., 2014), whereas 18 patients from another study scored 55.2 (12.7) on the Harris Hip Score (Harris, 1969).

In the ANOVA significant differences were detected between the bilateral patients, unilateral patients and healthy controls for both the ipsilateral and contralateral leg for all parameters except for the foot progression angle (not displayed) and the external ankle dorsiflexion/plantarflexion moment (Supplementary Figure 1 and Supplementary Figure 2).

Posthoc tests showed that bilateral patients walked with more thorax lean towards the ipsilateral side, more pelvic tilt and less pelvic obliquity up, less hip extension as well as less knee extension during the stance phase of gait compared to healthy controls (Figure 1). The contralateral leg showed additionally less hip abduction during the swing phase of gait and more dorsiflexion in the ankle joint (Figure 2). The range of motion of both the hip and the knee are limited in the bilateral hip OA patients.

Unilateral patients showed a similar pattern as the bilateral patients for the ipsilateral leg compared to the healthy controls, with more thorax lean towards the ipsilateral side and less hip and knee extension (Figure 1). However also differences were detected: whereas the bilateral patients had less pelvic obliquity up and more pelvic tilt, the unilateral patients showed more thorax tilt and less hip abduction (swing phase). For the contralateral leg, the unilateral patients only showed more thorax tilt and less hip abduction in the swing phase compared to healthy controls (Figure 2).

The ipsilateral leg of bilateral patients did not show any differences compared to the ipsilateral leg of unilateral patients. The contralateral leg of bilateral patients showed less hip and knee extension compared to the contralateral leg of unilateral patients (Figure 1 and Figure 2).

Regarding the kinetic parameters, it was found that the ipsilateral leg of patients with bilateral and unilateral hip OA had lower internal hip rotation moments, as well as a lower external knee adduction moments in the second half of stance compared to healthy controls (Figure 3 and Figure 4). On the ipsilateral side, bilateral patients additionally showed a lower hip extension moment compared to the healthy control group. Only the contralateral leg of bilateral patients showed differences compared to healthy controls: lower internal hip rotation moments and a reduced knee extension moment during the second half of stance.

A closer look at both legs of the bilateral patients showed differences between the ipsilateral leg (leg to be operated on) and the contralateral leg (Figure 5). The most striking differences were the counteracting movements of the thorax and the pelvis in the sagittal plane (tilt) and the reduced hip extension at the end of the stance phase of the ipsilateral leg. The lower hip flexion at the beginning of stance and the lower hip extension at the end of stance defined the reduced hip range of motion in the ipsilateral leg. The reduced range of motion resulted in a slightly lower external hip extension moment in the second half of stance phase for the ipsilateral leg.

Besides the above-mentioned differences (thorax tilt, pelvic tilt, hip flexion-extension and hip flexion extension moment), unilateral patients had an asymmetric thorax obliquity (more thorax lean towards the ipsilateral side), higher hip adduction moments during the first half of stance and higher knee adduction moments during the second half of stance in the contralateral leg (Figure 6). The contralateral leg further showed higher rotation moments and more knee extension during the second half of stance compared to the ipsilateral leg.

The symmetry values differed between bilateral patients and healthy controls only with regard to the hip moment impulse: healthy controls showed a higher gait symmetry compared to the bilateral patients (Figure 7). Unilateral patients showed a lower symmetry compared to healthy controls for the knee adduction moment during the second half of stance, the hip adduction moment during the first half of stance and the hip moment impulse. Only the symmetry in the knee adduction moment during the second half of stance differed significantly between the bilateral and unilateral patients, with a lower symmetry for unilateral patients.

The present study showed that bilateral patients walk with a gait pattern, which has been previously published for unilateral patients: more thorax tilt and lean, more pelvic tilt, less hip extension and less knee extension (Shakoor et al., 2003; Foucher et al., 2007; van Drongelen et al., 2021). The above-mentioned results were found for both legs of the bilateral hip OA patients. Kubota et al. (Kubota et al., 2007) also studied bilateral patients. They found that bilateral patients walked with a forward tilted pelvis and a lean to the ipsilateral side during stance, as was found in the present work. In the current study, significant differences were found between the ipsi- and contralateral leg of bilateral affected OA patients. The patients showed an increased pelvic tilt and a decreased hip extension towards the end of the stance phase for the ipsilateral leg. The contralateral leg had to counter the pelvic movement and showed more hip extension. A comparison to the study of Kubota et al. (Kubota et al., 2007) was not possible, as from that study it did not become clear whether mean values for both legs or data for just one leg were analyzed.

In the present study, a significant higher BMI was found for the unilateral patients compared to the healthy controls. BMI has been associated with lower knee and hip range of motion (Holla et al., 2011) and could be responsible for the differences between the unilateral patients and healthy controls. However, knee and hip range of motion have also been associated with disability (Steultjens et al., 2000) and since no differences were found between bilateral and unilateral patients in BMI nor in the knee and hip flexion the differences in BMI were neglected.

The second aim of the present study dealt with the magnitude and symmetry of the external knee and hip adduction moments. We hypothesized that, in contrast to patients with unilateral hip OA, the external knee and hip adduction moments of the contralateral side are not different compared to the ipsilateral side (more affected side) in bilateral affected patients with hip OA. As a result, bilateral hip OA patients show a more symmetrical joint load compared to unilateral hip OA patients.

For the hip, no differences in the external hip adduction moments were found between patients and healthy controls. This is in contrast to the findings of Foucher et al. (Foucher et al., 2007), who attributed the reduced hip adduction moments on the ipsilateral side in unilateral hip OA patients to an impairment of the hip abductor muscles. Also Kubota et al. (Kubota et al., 2007) reported lower external hip adduction moments. In that study, the reduction in the hip adduction moment was compensated with increased power generation in the ankle joint. However, the patients in the study of Kubota et al. (Kubota et al., 2007) had a very low walking speed (∼0.65 ms⁻¹) so that the characteristic double peak was not visible in the hip adduction moment, which might indicate that patients with low functional status were studied (Bahl et al., 2020). The patients in the present study walked at a speed similar to unilateral hip OA patients reported in various studies (Foucher, 2017; Schmidt et al., 2017). In comparison to the healthy controls, the typical double peak in the knee and hip adduction moments seemed to be missing as well (Figures 3,4). However, in the direct comparison of the legs (Figure 5) and in the comparison to unilateral patients it can be seen that the double peak is less pronounced and the curve flattened in patients with bilateral hip OA compared to healthy controls.

Besides the external hip adduction moment, the external knee adduction moments are important parameters (which can be determined by gait analysis), as the peak knee adduction moment has been related to the initiation and progression of knee OA (Miyazaki et al., 2002; Andriacchi and Mündermann, 2006). Both the bilateral and unilateral patients showed a reduced knee adduction moment in the second half of stance phase in the ipsilateral leg, but not in the contralateral leg, compared to the healthy controls. Whereas the unilateral patients showed a clear higher knee adduction moment in the contralateral knee, the bilateral patients had similar knee adduction moments in both knees.

The ipsilateral trunk lean toward the affected stance limb, shown in our patients with unilateral and bilateral hip OA, has been identified as an important compensatory mechanism to unload the hip (Stief et al., 2014). This compensating strategy is also responsible for the redistribution of loads transferred through the medial and lateral knee compartment (Simic et al., 2012; Stief et al., 2014). Body weight and thus the ground reaction force vector is moved toward the center of the knee and hip joint, which reduces medial knee load in a dose-response relationship, with larger lean angles leading to greater reductions in the knee adduction moment (Simic et al., 2012). Mündermann et al. (Mündermann et al., 2008) and van den Noort et al. (van den Noort et al., 2013) have demonstrated that lateral trunk lean has the potential of reducing the knee adduction moment during walking by up to 65% in healthy subjects. Besides the unloading mechanism, ipsilateral trunk lean can shift the physiological external knee adduction moment into a pathological external abduction moment (valgus moment). This lateral adjustment is argued to increase the lateral tibio-femoral compartment load (Sharma et al., 2000). However, at the moment, it is still questionable whether the change in the medial-to-lateral knee joint load distribution in the ipsilateral leg leads to increased progression or even initiates degenerative changes of the lateral compartment of the knee, since lower medial loading does not necessarily indicate higher lateral loading (van Drongelen et al., 2020b). Various studies reported a higher risk for the development of OA in the contralateral knee joint after THA (Shakoor et al., 2003; Umeda et al., 2009). However, these studies only reported degenerative findings and did not specify the medial or lateral knee compartment. Gillam et al. (2013) stated that further studies are needed to investigate whether other factors besides load play a role as well. At the moment only Weidow et al. (Weidow et al., 2005) showed an increased probability of OA in the lateral knee compartment of the ipsilateral leg in patients with hip OA which could be attributed to the morphology of the hip and pelvis in the patient collective.

We hypothesized that bilateral hip OA patients show a more symmetrical joint load compared to unilateral hip OA patients due to the symmetrical joint moments. Indeed, bilateral hip OA patients walked with more symmetrical knee and hip joint moments, almost similar to healthy controls. Compared to the healthy controls, the bilateral patients only showed lower symmetry in the hip moment impulse. Besides the asymmetry in the hip moment impulse, the unilateral patients also showed a lower symmetry for the hip adduction moment in the first half of stance and the knee adduction moment in the second half of stance compared to the healthy controls.

A closer look at the symmetry data showed that some patients were outside the whiskers, which represent 1.5 times the interquartile range (outliers not displayed in Figure 7). However, in general the gait symmetry of bilateral hip OA patients is comparable to the healthy controls. The present study confirmed the values presented by Schmidt et al. (Schmidt et al., 2017) who found clearly lower symmetry for unilateral hip OA patients compared to healthy controls. Schmidt et al. (Schmidt et al., 2017) reported more variation for unilateral hip OA patients with regard to ASI than healthy controls. Individual adopted gait patterns, which might be caused by pain, can explain the inter-subject variability.

With a more symmetrical load, one could argue that the overload on the contralateral leg is not present in bilateral patients, which could lower the risk for knee and hip OA in the contralateral leg. It must be noted that ASI does not give any qualitative information on which side is overloaded or underloaded. Further, the knee adduction moments in bilateral hip OA patients were lower in both legs compared to healthy controls (significant for the ipsilateral leg and not significant for the contralateral leg). As mentioned above, a lower knee adduction moment might indicate a shift of the load from the medial to the lateral knee compartment (Schmidt et al., 2017), which can facilitate degenerative changes of the lateral knee compartment.

A direct comparison of bilateral and unilateral hip OA patients is worth discussing. Usually, bilateral affected patients are excluded from studies on hip OA, which reduces the sample size considerably. From the present study, it became clear that there were no differences in the kinematics and kinetics of the ipsilateral leg between unilateral and bilateral hip OA patients. As such, it can be concluded that the ipsilateral leg of the bilateral patients could be included in studies next to unilateral hip OA patients. In contrast, and based on the larger hip range of motion, the contralateral leg of bilateral patients is not automatically suited for inclusion in studies like the ipsilateral leg.

External knee and hip adduction moments were chosen as the primary outcome, as they are still the most commonly used measures of joint loading with an important relationship to disease progression and initiation (Hurwitz et al., 2001; Miyazaki et al., 2002; Andriacchi and Mündermann, 2006; Lenaerts et al., 2009; Wesseling et al., 2015; Tateuchi et al., 2017). At the moment it is still questionable whether the change in medial-to-lateral knee joint load distribution in the ipsilateral leg leads to increased progression or even initiates degenerative changes of the lateral compartment of the knee, since lower medial loading does not necessarily indicate higher lateral loading (Andriacchi, 2013). Alternative measures of joint load, including compressive force in the medial and lateral knee compartment measured via by musculoskeletal modeling including muscle activity and forces (Holder et al., 2020), might be able to answer this question in the future.